Christoffer Lerno
Christoffer Lerno
There should never be instructions with the same name but different number of arguments. That is what the asm instruction variant is for.
In the LDR case, isn't that just an address? So it's not different arity in the C3 asm version, because it's treated as a memory location.
The add instruction with lsl doesn't fit the grammar any, so the actual way to do that is: `add $x0, $x1 + $x2
So, it's possible it's needed to patch this, but let's first find the case where it's needed.
It should support `add $x0, $x1 + $x2
It doesn't though, and that is a bug. So please file a bug for that and any similar instructions that should support it.
Not sure what you need it for. Cross product on 3d vectors already of course. Feel free to contribute these operations if you need them.
I am deeply suspicious of automatically pulling in additional dependencies. That's what making development go to ****. Because if there is no *friction* to pulling in a billion dependencies, then...
This has led to some discussion. What's done here is odd: we're running "bench" which takes any target and tries to run its benchmarks, but in this case the target...
But that means you'd have a lot of targets just for testing. I am not sure this is right.